The broad goals of this project are to understand the computations that can be performed by a neural circuit. The vertebrate retina is chosen as a model system, because of our ability to achieve complete recordings of its output while stimulating with realistic patterns of light, as well as its intricate and well-characterized anatomy. The retina's representation of visual stimuli changes dramatically between the photoreceptors, which use a camera-like representation where each cell is a pixel, and the ganglion cells, which represent the image with a complex and highly overlapping basis set of visual features. We seek to characterize the manner in which visual scenes are represented by the complete population of retinal ganglion cells. The Specific Aims are: 1) characterize the diversity and stereotypy of the visual features represented within the population of ganglion cells by performing functional classification;2) measure the correlation both within and among the parallel visual channels;3) formulate decoding strategies to perform visual discrimination using large populations of ganglion cells. Our functional classification will minimize the effects of subtle variations from one retinal preparation to the next by recording simultaneously from 200+ ganglion cells from a single retina and will use information theory to assess functional differences using a battery of visual stimuli including natural movies. Our characterization of the patterns of correlation among groups of ganglion cells will use new information-theoretic techniques, including the maximum entropy approach. Finally, our decoders will seek to make use of our models of correlation in large populations to improve the accuracy of visual discriminations.